optimal m bius transformation for information
play

OptimalMbiusTransformation forInformationVisualizationandMeshing - PowerPoint PPT Presentation

OptimalMbiusTransformation D.Eppstein,UCIrvine,WADS2001 OptimalMbiusTransformation forInformationVisualizationandMeshing MarshallBern XeroxPaloAltoResearchCtr. DavidEppstein


  1. Optimal฀Möbius฀Transformation D.฀Eppstein,฀UC฀Irvine,฀WADS฀2001 Optimal฀Möbius฀Transformation for฀Information฀Visualization฀and฀Meshing Marshall฀Bern Xerox฀Palo฀Alto฀Research฀Ctr. David฀Eppstein Univ.฀of฀California,฀Irvine Dept.฀of฀Information฀and฀Computer฀Science

  2. Optimal฀Möbius฀Transformation D.฀Eppstein,฀UC฀Irvine,฀WADS฀2001 What฀are฀Möbius฀transformations? Fractional฀linear฀transformations฀of฀complex฀numbers: z ฀ → ฀( a฀z ฀+฀ b )฀/฀( c฀z ฀+฀ d ) But฀what฀does฀it฀mean฀geometrically? How฀to฀generalize฀to฀higher฀dimensions? What฀is฀it฀good฀for?

  3. Optimal฀Möbius฀Transformation D.฀Eppstein,฀UC฀Irvine,฀WADS฀2001 What฀are฀Möbius฀transformations?฀฀(II) Circle-preserving฀maps฀from฀the฀plane฀to฀itself More฀intuitive Generalizes฀nicely฀to฀spheres,฀higher฀dimensional฀spaces Not฀very฀concrete

  4. Optimal฀Möbius฀Transformation D.฀Eppstein,฀UC฀Irvine,฀WADS฀2001 What฀are฀Möbius฀transformations฀฀(III) Inversion:฀map฀radii฀of฀circle฀to฀same฀ray so฀that฀product฀of฀distances฀from฀center฀=฀radius 2 Möbius฀transformation฀=฀composition฀of฀multiple฀inversions More฀concrete,฀still฀generalizes฀nicely

  5. Optimal฀Möbius฀Transformation D.฀Eppstein,฀UC฀Irvine,฀WADS฀2001 What฀are฀Möbius฀transformations?฀฀(IV) View฀plane฀as฀boundary฀of฀halfspace฀model฀of฀hyperbolic฀space Möbius฀transformations฀of฀plane฀ ↔ ฀hyperbolic฀isometries Esoteric Most฀useful฀for฀our฀algorithms

  6. Optimal฀Möbius฀Transformation D.฀Eppstein,฀UC฀Irvine,฀WADS฀2001 Optimal฀Möbius฀transformation: Given฀a฀planar฀(or฀higher฀dimensional)฀input฀confjguration Select฀a฀Möbius฀transformation from฀the฀(six-dimensional฀or฀higher)฀space฀of฀all฀Möbius฀transformations That฀optimizes฀the฀shape฀of฀the฀transformed฀input Typically฀min-max฀or฀max-min฀problems: maximize฀min(set฀of฀functions฀describing฀transformed฀shape฀quality)

  7. Optimal฀Möbius฀Transformation D.฀Eppstein,฀UC฀Irvine,฀WADS฀2001 Application:฀spherical฀graph฀drawing Theorem฀[Koebe,฀Thurston]:฀Any฀planar฀graph฀can฀be฀represented by฀disjoint฀disks฀on฀a฀sphere฀so฀two฀vertices฀adjacent฀iff฀two฀disks฀tangent For฀maximal฀planar฀graphs,฀unique฀up฀to฀Möbius฀transformation Other฀graphs฀can฀be฀made฀maximal฀planar฀by฀adding฀vertex฀in฀each฀face Optimization฀problem: Find฀disk฀representation฀of฀ G ฀maximizing฀minimum฀disk฀radius or,฀given฀one฀disk฀representation,fjnd฀Möbius฀transformation฀maximizing฀min฀radius Solution฀also฀turns฀out฀to฀display฀all฀symmetries฀of฀initial฀embedded฀graph

  8. Optimal฀Möbius฀Transformation D.฀Eppstein,฀UC฀Irvine,฀WADS฀2001 Application:฀conformal฀mesh฀generation Given฀simply-connected฀planar฀domain฀to฀be฀meshed Map฀to฀square,฀use฀regular฀mesh,฀invert฀map฀to฀give฀mesh฀in฀original฀domain Different฀points฀of฀domain฀may฀have฀different฀requirements฀for฀element฀size To฀minimize฀#฀elements,฀map฀regions฀requiring฀small฀size฀to฀large฀areas฀of฀square Conformal฀map฀is฀unique฀up฀to฀Möbius฀transformation Optimization฀Problem: Find฀conformal฀map฀maximizing฀min(size฀requirement฀*฀local฀expansion฀factor) to฀minimize฀overall฀number฀of฀elements฀produced

  9. Optimal฀Möbius฀Transformation D.฀Eppstein,฀UC฀Irvine,฀WADS฀2001 Application:฀hyperbolic฀browser ฀[Lamping,฀Rao,฀and฀Pirolli,฀1995] Technique฀for฀viewing฀large฀web฀sites฀or฀other฀structured฀information by฀laying฀out฀information฀in฀hyperbolic฀space Allows฀“fjsheye฀view”:฀close-up฀look฀at฀details฀of฀some฀point฀in฀site global฀structure฀of฀site฀visible฀towards฀boundary฀of฀hyperbolic฀model The฀farther฀away฀a฀point฀is฀from฀the฀viewpoint฀(in฀hyperbolic฀distance) the฀smaller฀the฀information฀it฀represents฀will฀be฀displayed Optimization฀problem: Find฀good฀initial฀viewpoint฀for฀hyperbolic฀browser in฀order฀to฀make฀overall฀site฀as฀visible฀as฀possible Maximize฀minimum฀size฀of฀displayed฀object or Maximize฀minimum฀separation฀between฀two฀objects

  10. Optimal฀Möbius฀Transformation D.฀Eppstein,฀UC฀Irvine,฀WADS฀2001 Application:฀brain฀fmat฀mapping฀ [Hurdal฀et฀al.฀1999] Problem:฀visualize฀the฀human฀brain All฀the฀interesting฀stuff฀is฀on฀the฀surface But฀diffjcult฀because฀the฀surface฀has฀complicated฀folds Approach:฀fjnd฀quasi-conformal฀mapping฀brain฀ → ฀plane Then฀can฀visualize฀brain฀functional฀units฀as฀regions฀of฀mapped฀plane Avoids฀distorting฀angles฀but฀areas฀can฀be฀greatly฀distorted As฀in฀mesh฀gen.฀problem,฀mapping฀unique฀up฀to฀Möbius฀transformation Optimization฀problem: Given฀map฀3d฀triangulated฀surface฀ → ฀plane, fjnd฀Möbius฀transformation฀minimizing฀max(area฀distortion฀of฀triangle)

  11. Optimal฀Möbius฀Transformation D.฀Eppstein,฀UC฀Irvine,฀WADS฀2001 Optimal฀Möbius฀Algorithm Key฀components: Quasiconvex฀Programming Hyperbolic฀Geometry

  12. Optimal฀Möbius฀Transformation D.฀Eppstein,฀UC฀Irvine,฀WADS฀2001 Quasiconvex฀functions Level฀sets฀are฀nested฀convex฀curves Inner฀curves฀correspond฀to฀smaller฀function฀values (Like฀topographic฀map฀of฀open฀pit฀mine)฀

  13. Optimal฀Möbius฀Transformation D.฀Eppstein,฀UC฀Irvine,฀WADS฀2001 Quasiconvex฀programming฀ [Amenta,฀Bern,฀Eppstein฀1999] Given฀ n ฀quasiconvex฀functions฀ f i max( f i ( x ))฀is฀also฀quasiconvex problem฀is฀simply฀to฀compute฀ x฀ minimizing฀max( f i ( x )) Can฀be฀solved฀exactly฀with฀O( n )฀constant-size฀subproblems using฀low-dimensional฀linear-programming-type฀techniques Can฀be฀solved฀numerically฀by฀hill-climbing฀or฀other฀local฀optimization฀methods

  14. Optimal฀Möbius฀Transformation D.฀Eppstein,฀UC฀Irvine,฀WADS฀2001 Hyperbolic฀space฀(Poincaré฀model) Interior฀of฀unit฀sphere;฀lines฀and฀planes฀are฀spherical฀patches฀perpendicular฀to฀unit฀sphere

  15. Optimal฀Möbius฀Transformation D.฀Eppstein,฀UC฀Irvine,฀WADS฀2001 Two฀models฀of฀hyperbolic฀space Klein฀Model Hyperbolic฀objects฀are฀straight฀or฀convex iff฀their฀model฀is฀straight฀or฀convex Angles฀are฀severely฀distorted Hyperbolic฀symmetries฀are฀modeled฀as Euclidean฀projective฀transformations Poincaré฀Model Angles฀in฀hyperbolic฀space equal฀Euclidean฀angles฀of฀their฀models Straightness/convexity฀distorted Hyperbolic฀symmetries฀are฀modeled฀as Möbius฀transformations

  16. Optimal฀Möbius฀Transformation D.฀Eppstein,฀UC฀Irvine,฀WADS฀2001 ฀Möbius฀transformation฀and฀hyperbolic฀geometry Quasiconvex฀programming฀works฀equally฀well฀in฀hyperbolic฀space Due฀to฀convexity-preserving฀properties฀of฀Klein฀model But฀space฀of฀Möbius฀transformations฀is฀not฀hyperbolic฀space... View฀Möbius฀transformation฀as฀choice฀of฀Poincaré฀model Factor฀transformations฀into choice฀of฀center฀point฀in฀hyperbolic฀model฀(affects฀shape) Euclidean฀rotation฀around฀center฀point฀(doesn’t฀affect฀shape)

  17. Optimal฀Möbius฀Transformation D.฀Eppstein,฀UC฀Irvine,฀WADS฀2001 Optimal฀Möbius฀transformation฀algorithm Represent฀optimization฀problem฀objective฀function as฀max฀of฀set฀of฀quasiconvex฀functions where฀function฀argument฀is฀hyperbolic฀center฀point฀location Hard฀part:฀proving฀that฀our฀objective฀functions฀are฀quasiconvex Solve฀quasiconvex฀program Use฀center฀point฀location฀to฀fjnd฀Möbius฀transformation

  18. Optimal฀Möbius฀Transformation D.฀Eppstein,฀UC฀Irvine,฀WADS฀2001 Conclusions Formulate฀several฀interesting฀applications฀as฀Möbius฀optimization Can฀solve฀via฀LP-type฀techniques฀or฀hill-climbing Interesting฀use฀of฀hyperbolic฀methods฀in฀computational฀geometry but... Details฀of฀exact฀algorithm฀may฀be฀diffjcult฀to฀implement (see฀Gärtner฀for฀similar฀diffjculties฀in฀LP-type฀min-volume฀ellipsoid) Not฀able฀to฀prove฀quasiconvexity฀in฀some฀cases e.g.฀given฀number฀x,฀triangle฀T฀at฀infjnity฀in฀hyperbolic฀3-space are฀points฀from฀which฀T฀subtends฀solid฀angle฀>฀x฀convex?

Download Presentation
Download Policy: The content available on the website is offered to you 'AS IS' for your personal information and use only. It cannot be commercialized, licensed, or distributed on other websites without prior consent from the author. To download a presentation, simply click this link. If you encounter any difficulties during the download process, it's possible that the publisher has removed the file from their server.

Recommend


More recommend